China’s statistical conundrums – the example of the PMIs

September 12th, 2018 by Hubert Fromlet, Kalmar

Surprisingly, one still can find people who believe that statistical quality in China has been improving in recent years. I myself cannot find evidence for such a conclusion, neither can our LNU China Panel (https://blogg.lnu.se/china-research/files/2018/05/ChinaPanelSurvey-May-2018.pdf ; at least there is no marketing or concrete information on such a desirable change coming from Chinese authorities. Still, there is a lot to wonder about.

One discrepancy that still makes me puzzled concerns statistics on GDP, industrial production and the Purchasing Manager Indices (PMIs). It may be added that I feel quite sure to understand the technical mechanisms of PMIs since I myself prepared and introduced the Swedish PMI almost 25 years ago together with my colleague Åke Gustafsson from Swedbank. Sometimes, statistical correlation between these micro and macro indicators is questioned ; and I myself – following my own historical experience – cannot either see sufficient consistency right now between the Chinese PMIs and Chinese industrial production or GDP – possibly or probably as a result of insufficient statistical quality rather than non-existing correlation.

Two PMIs in China

Let’s look at some numbers. There are actually two PMIs in China. One is produced by the National Bureau of Statistics (NBS) and the China Federation of Logistics and Purchases (CFLP) with focus on 3000-4000 larger companies including SoEs, the other one by Caixin/Markit with 430 private, mainly medium-sized and smaller companies. According to statistics for August 2018, the NBS/CFLP PMI for manufacturing rose slightly from 51.2 to 51.3 compared to the month before, whereas the corresponding Caixin/Markit index fell to 50.6 from 50.8. These numbers – not very far from the “borderline” of 50 between better and weakening growth – have been in these numerical regions for quite some time.

Abroad, the Caixin/Markit PMI tends to receive better recognition than the NBS/CFLP index because of no or only limited governmental influence. However, both indices are interpreted and commented very strictly when it comes to the borderline of 50. Just above 50 is usually regarded as positive and just below 50 as negative. But users of the PMI indices should be aware of the fact that both PMI series are calculated as diffusion indices that do not really reflect the strength of changes in the participating individual companies – and, consequently, only the direction. For this reason, I cannot warn enough for putting too much positive focus on numbers slightly above/below 50 and too encouraging/discouraging oral and written comments on further expansion or contraction. Some months of observation and/or moving averages could be useful.

It should be kept in mind that the Caixin /Markit PMI seems to be a notch closer to industrial reality than the official NBS/CFLP index – and, consequently, better capturing the currently fading export performance of Chinese industry.

Do Chinese PMIs reflect reality?

My view is that China’s PMI numbers during the past quarters have come in on (somewhat) too high levels. Again, a number not too much below 50 does not mean that a recession is going on – instead, for some time, only a declining rate of growth; and a number slightly above 50 does not necessarily point at an further improving industrial activity any time soon.

My strong feeling is that the real state of the Chinese industry seems to be (somewhat?) weaker these days than recent PMI indicators for manufacturing slightly above 50 may indicate. American protectionism hurts.

However, lagging transparency may mean that not all my conclusions necessarily are correct. One should be cautious and humble when interpreting Chinese statistics, also those for the PMIs.

Anyway, the globe’s second largest economy, China, still has a lot of institutional homework to do, improvements of national statistics included.

Hubert Fromlet
Affiliate Professor at the School of Business and Economics, Linnaeus University
Editorial board

 

Back to Start Page

Comments are closed.